Qdrant Advanced Collection Configuration Named Vectors (Multi-Vector per Point) Store multiple embeddings per document — e.g., title, body, image — search any of them or combine. Sparse Vectors (SPLADE / BM25-style) Hybrid Search with Fusion (Qdrant 1.10+) Qdrant performs server-side RRF / DBSF fusion across named + sparse vectors in a single call. ColBERT-style Late Interaction with Rescoring Quantization Options | Type | Memory reduction | Speed | Recall impact | |---|---|---|---| | Scalar int8 | 4x | 2x faster | -1 to -2% | | Binary | 32x | 40x faster | -5 to -15% (recover with rescoring)…